Data & AI

Progress Agentic RAG ranked #1 on Snowflake's MADQA Leaderboard, an independent benchmark built with Oxford, UNC-Chapel Hill, and Hugging Face that tests whether AI answers are accurate and properly grounded in real documents, not just plausible-sounding. The win validates a year of work since the Nuclia acquisition building the Agentic Knowledge Layer (agentic capability embedded across ingestion, retrieval, and evaluation) and lands as the broader market shifts toward trust and governance as the deciding factor for enterprise AI.

Traditional RAG systems are designed around a simple idea: find the most relevant information, add it to the prompt and generate an answer. This works well for questions where the answer can be found in a small set of relevant documents.

An agentic workflow is a multi-step process where an AI agent plans the steps, picks tools, acts and adjusts based on what comes back, instead of following a script written in advance. In this post, we walk through how that loop runs, where it beats traditional automation, and the guardrails it needs in production.

Hybrid search combines keyword and semantic retrieval so one retriever handles both exact tokens and reworded concepts, not either alone.